Catching Unicorns with GLTR logo

Catching Unicorns with GLTR

Paid

Spot AI-generated text at a glance with GLTR’s color-coded statistical forensics.

#AI Detection#Text Analysis#Tool#Open-Source#MIT-IBM#HarvardNLP#Machine-generated Text#Forensic
Inputs: text
Type
Saas

About Catching Unicorns with GLTR

MailChimpMailChimp is an easy-to-use email marketing platform designed to help businesses of all sizes create, send, and track professional email campaigns. With MailChimp, you can design beautiful, eye-catching emails in minutes with their intuitive drag-and-drop editor, or import existing HTML templates. You can also personalize your emails with customizable fields and send them to targeted audiences. Plus, MailChimp’s powerful analytics tools give you insights into the performance of your emails, so you can track opens, clicks, and unsubscribes. With MailChimp, you can get your message out to the right people, at the right time, and see the results of your efforts in real time. Make your email marketing efforts count with MailChimp.

Key Features

Color-coded token visualization (green/yellow/red/purple) by probability rank
Live demo to paste any text for instant analysis
Histograms showing distribution of top-10/top-100/top-1000 ranks
Entropy and predictability summaries for entire texts and sentences
Hover tooltips with top predicted words and ranks
Sentence-level breakdowns with per-sentence charts
Model selection for different GPT-2 variants
Pre-loaded real and fake text examples for comparison
Analysis examples of algorithmic journalism (e.g., sports reports)
Open-source code available on GitHub for local use and customization

Pros & Cons

Pros
  • Open-source and free to use
  • Visual color coding makes it easy to interpret word predictability at a glance
  • Proven to improve human detection accuracy from ~54% to ~72% (ACL 2019 study)
  • Live demo available for quick experimentation without installation
  • Hover tooltips reveal the model's top predictions, aiding forensic analysis
Cons
  • Relies on GPT-2 117M model; performance may vary with newer or larger models
  • Not a definitive classifier; results depend on text length and writing style
  • Skilled human writers or advanced AI can produce text that passes the test
  • Requires understanding of probability ranks and histograms for full benefit

Best For

Journalists: Vet suspicious articles or press releases for signs of machine-generated language.Educators: Check student assignments for AI-writing patterns while teaching responsible AI use.Editors: Screen submissions for synthetic text before fact-checking and copyediting.Researchers: Analyze outputs from language models to study predictability and generation artifacts.Policy/Compliance teams: Assess documents for AI content to support disclosure and transparency policies.OSINT/Investigators: Inspect online posts and comments for coordinated AI-generated messaging.Product teams: Evaluate in-app generative features and tune safety or disclosure UX with evidence.Content moderators: Triage suspected AI spam or bot posts with quick visual cues.Marketing analysts: Audit vendor-provided copy for potential undisclosed AI authorship.Educators/Workshop leads: Demonstrate AI text properties in classrooms and training sessions.

Alternatives to Catching Unicorns with GLTR